I=
Pakistan Research Repository Home
 

Title of Thesis

Expert System For Optimization Of Welding Process Of Thin Walled HSLA Steel Structures

Author(s)

Naeem Ullah Dar

Institute/University/Department Details
Department of Mechanical Engineering, Faculty of Mechanical and Aeronautical Engineering / University of Engineering & Technology, Taxila
Session
2009
Subject
Mechanical Engineering
Number of Pages
320
Keywords (Extracted from title, table of contents and abstract of thesis)
Expert, Full Factorial, Simulation, Tig Welding, Maximization, Minimization, Residual, Stresses, Process, Walled, Optimization, Weld Strength, Thin, Walled, Structures

Abstract
With the introduction of welding as joining method, the welding technology was  applied as major joining technique in hi-tech industries to the welding of steels for  manufacturing of different structures like pressure vessels and aerospace applications.Mostly high strength low alloy steels in thin cylindrical shell form are being used for  aerospace structures due to high strength and low weight ratio. Despite being high  strength and light weight by numerous advantages, the welding of thin walled structure of
high strength low alloy steel (also known as HSLA Steel) comes also with a major problems of weld induced imperfections due to high temperatures like residual stresses  and distortions with shortening of weld strength and it is a still major challenge for the  welding professionals due to the complex nature of the welding phenomenon despite  many innovations in welding technology. The most of the weld induced imperfections are  the result of transient temperature distributions and subsequent cooling of the welds
followed by transient and residual stress fields.
Where as, the reliability of thin-walled structures used for any aerospace or  pressure vessel application is on the prime importance every time for safe operational.Usually, thin walled cylindrical structures contain two types of weld as longitudinal and  circumferential. The major design and industry constraints are weld strength and cost  competitive. Gas Tungsten Arc Welding (GTAW) or TIG process is mostly applied due  to the excellent weld strength and cost competitiveness.The main aim of this research work is to analyze and experimentally investigate the TIG welding parameters for purpose of minimizing residual stresses and distortion with the requirements of maximizing of weld strength of thin walled structures of HSLA steel respectively. To achieve the aforementioned targets, the following strategy was applied keeping in view the complex phenomena of welding, time and cost of extensive experimentations involved.
Weld experiments were subdivided into linear and circumferential welding.Initially for linear welding, TIG welding parameters were analyzed to determine their significance on thin plates of HSLA steel of different thicknesses (3 to 5 mm) by following design of experiments (DOE) with employing 2-level full factorial and response surface method (RSM) designs to have response (weld strength, distortion & residual stress). Whereas for circumferential welding, a hybrid numerical simulation and experimental based analysis approach was employed to model and predict TIG welding process to investigate the transient temperature distributions, transient/residual stress fields and distortion for circumferentially welded thin-walled cylinders of HSLA steel.The simulations strategy was developed and implemented by using commercial available general purpose finite element software ANSYS® enhanced with subroutines. First thermal analysis was completed followed by a separate mechanical analysis based on the thermal history. From the three dimensional FE model developed for TIG welding process of circumferential welding, a series of virtual welding experiments based on statistical designs (DOE) were performed for response (residual stresses and distortion) with different thicknesses by using full factorial and RSM as applied for linear welding.
The effects of following six parameters, four numeric and two categorical: welding current, welding voltage, welding speed, sheet/cylinder thickness and trailing (Ar) & weld type (linear and circumferential) were investigated upon following three performance measures: weld strength, residual stresses and distortions for different thicknesses of material of HSLA steel. The experimental results were analyzed using ANOVA and significance of effects of all the tested parameters upon performance measures was determined. Empirical models for weld strength, distortion and residual stresses, in terms of significant parameters, were also developed and numerical optimization was performed according to the desirability for the maximization of weld strength and minimization of distortion & residual stresses. All the statistical analyses were performed by using commercial available statistical software Design-Expert® and MINITAB
From the results of post-experimental analyses, it was noticed that the effects of welding current, welding voltage and welding speed upon weld strength, residual stresses and distortion are extremely significant, while the effect of trailing and weld type is also considerably significant with respect to material thicknesses. The residual stresses are highly sensitive to heat input (weld temperatures). The residual stresses and distortion in circumferential welding are low as compared to linear welding for the same welding parameters and material thickness respectively. The vital recommendation, in this regard, is to use the parameters of welding resulting low input heat (low current, low voltage and high speed) with application of trailing with respect to material thicknesses for the maximum weld strength and minimum residual stresses and distortion in thin walled structures of HSLA steel.For the trade-off among aforementioned opposing targets and for prediction of values of performance measures at different settings of TIG welding parameters, the expert system tool, employing fuzzy reasoning mechanism, was utilized. Initially, an expert system was developed for the optimization of parameters according to objectives of maximization and/or minimization of weld strength, distortion and residual stresses.The expert system also provided the predicted values of various performance measures based upon the finalized values of the welding parameters. The analyses, simulations, experimental and ANOVA results were utilized for the making of fuzzy rule-base.The fuzzy rule-base was adjusted for maximum accuracy by employing the simulated annealing (SA) algorithm.
In the next stage, a machine learning (ML) technique was utilized for creation of a expert system, named as EXWeldHSLASteel, that can: self-retrieve and self-store the experimental data; automatically develop fuzzy sets for numeric variables involved; automatically generate rules for optimization and prediction rule-bases; resolve the conflict among contradictory rules; and automatically update the interface of expert system according to newly introduced TIG welding process variables. The algorithms for these constituents were coded using a pointer-enabled language in C++. The coding involves a data structure named as doubly linked list, which provide the means for fast and efficient processing.The presented expert system is used for deciding the values of important welding process parameters as per objective before the start of actual welding process on shop floor. The user should be absolutely clear about the nature and requirements of any given TIG welding process, e.g., the setting parameters, fixed parameters, and geometric parameters etc. The expert system developed in the domain of welding for optimizing welding process of thin walled HSLA steel structure possesses all capabilities to adapt effectively to the unpredictable and continuously changing industrial environment of mechanical fabrication and manufacturing and to serve the newly emerging field of knowledge management by transforming individual (expert) knowledge into organizational knowledge i.e. implicit to explicit knowledge.

Download Full Thesis
4,856 KB
S. No. Chapter Title of the Chapters Page Size (KB)
1 0 CONTENTS

 

vii
 52 KB
2

1

INTRODUCTION

1.1 Welding Technolog
1.2 GTAW or TIG Welding
1.3 Variables and Performance Measures in GTAW Process
1.4 Current Challenges in Welding Domain
1.5 Application of Artificial Intelligence and Expert System to Manufacturing
1.6 Objectives, Research Methodology and Organization of Dissertation

1
456 KB
3 2 LITERATURE REVIEW

2.1 Introduction
2.2 Literature Survey in Welding Domain
2.3 Welding Induced Residual Stresses Measurement
2.4 Literature Survey in Artificial Intelligence and Expert System to Manufacturing
2.5 Chapter Summary and Conclusions

21
238 KB
4 3 ANALYZING & OPTIMIZING TIG WELDING PROCESS PARAMETERS

3.1 Introduction
3.2 Design of Experiments
3.3 Experimental Setup
3.4 Experimental Results, ANOVA, Regression and Optimization 
3.5 Numerical Optimization and Emperical Modeling using Response Surface Method (RSM)
3.6 Chapter Summary and Conclusions

41
3,100 KB
5 4 FE MODELING & SIMULATION OF GTAW PROCESS OF THIN WALLED STRUCTURE FOR CIRCUMFERENTIAL WELDING

4.1 Introduction
4.2 FE Modeling & Simulation Methodology
4.3 Welding Induced Stresses and Distortions
4.4 Experimental Setup for Validation of FE Models
4.5 Chapter Summary and Conclusions

99
1,642 KB
6 5 VIRTUAL DESIGN OF EXPERIMENTS (DOE) & OPTIMIZATION OF GTAW PROCESS OF THIN- WALLED STRUCTURE FOR CIRCUMFERENTIAL WELDING

5.1 Introduction
5.2 Analyzing the Effects of Welding Parameters on Residual Stresses
5.3 Virtual Design of Experiments (DOE) and Optimization of Circumferential Welding
5.4 Virtual Experiments Results, ANOVA, Regression, and Optimization
5.5 Linear & Circumferential Welding Optimization using RSM
5.6 Chapter Summary and Conclusions

141
2,722 KB
7 6 KNOWLEDGE ENGINEERING FOR OPTIMIZING TIG WELDING PROCESS

6.1 Introduction
6.2 The Objectives of Expert System and Application to Welding
6.3 Expert System Configuration
6.4 Fuzzy Reasoning for the Expert System
6.5 Optimal Formation of Rule-Base
6.6 Application Example
6.7 Chapter Summary and Conclusions

207
230 KB
8 7 THE SELF-DEVELOPING EXPERT SYSTEM FOR OPTIMIZING TIG WELDING PROCESS

7.1 Introduction
7.2 Self-Development of Expert System
7.3 Data Acquisition Module
7.4 Self-Development of Fuzzy Sets Module
7.5 Self-Development of Prediction Rule-Base Module
7.6 Self-Development of Optimization Rule-Base Module
7.7 Data Structures and Coding
7.8 Application Examples
7.9 Chapter Summary and Conclusions

223
905 KB
9 8 CONCLUSIONS & RECOMMENDATIONS

8.1 Conclusions
8.2 The Recommendations

245
86 KB
10 9 REFERENCES AND APPENDIX

 

251
290 KB